Generalized and Efficient Non-Line-of-Sight Imaging

Pablo Luesia-Lahoz, under supervision of Adolfo Muñoz and Diego Gutierrez

Abstract: #

Non-Line-of-Sight (NLOS) imaging aims to see what is hidden around a corner, with a variety of compelling applications such as medical imaging, search and rescue operations, or car safety. The field emerged with the advent of transient imaging, which, utilizing an extremely high temporal resolution, offers a means to analyze light transport over time. By capturing indirect photons reflected from a visible \emph{relay wall} and exploiting their time of flight, NLOS methods can reconstruct the geometry of hidden scenes.

Despite recent promising results, NLOS imaging is confined to laboratory-controlled situations, limiting its practical deployment. This thesis aims to progress NLOS imaging towards its practical implementation into everyday technology. To that end, we contribute in three different aspects that shape the structure of this thesis: generalization of NLOS for varying and complex setups, efficiency of reconstruction methods, and correct modeling of its transient capture.

Documents #

Thesis Document: PDF

Presentation: PPTX